Aquesta fotografia és de l'artista Michael Najjar, i en cert sentit és real, doncs l'artista va anar a Argentina a fer la foto. Però també és ficció. Hi ha un munt de feina al darrere. El que l'artista ha fet, en realitat, és modificar, digitalment, tots els contorns de les muntanyes per a fer-les encaixar amb les fluctuacions de l'índex Dow Jones Així que, això que veieu, aquest gran precipici al costat de la vall, és la crisi financera del 2008. La foto es va fer quan érem a les profunditats de la vall, allà. No sé on som ara. Això és l'índex Hang Seng de Hong Kong. Té una topografia similar, i em pregunto perquè.
This is a photograph by the artist Michael Najjar, and it's real, in the sense that he went there to Argentina to take the photo. But it's also a fiction. There's a lot of work that went into it after that. And what he's done is he's actually reshaped, digitally, all of the contours of the mountains to follow the vicissitudes of the Dow Jones index. So what you see, that precipice, that high precipice with the valley, is the 2008 financial crisis. The photo was made when we were deep in the valley over there. I don't know where we are now. This is the Hang Seng index for Hong Kong. And similar topography. I wonder why.
I això és art, és metàfora. Però el més important, penso jo, és que és una metàfora amb força. I és amb aquesta força que vull proposar, avui, que pensem una mica en el paper de les matemàtiques modernes -- no només matemàtiques financeres, sinó matemàtiques en general En la seva transició de quelcom que extraiem i derivem del món cap a quelcom que, en realitat, comença a donar-li forma -- tant al món que ens envolta com al nostre món interior. En particular, els algoritmes són les matemàtiques que els ordinadors usen per prendre decisions. Adquireixen sensibilitat a la veritat a base de repetir una vegada rere una altra. I es van ossificant i calcificant i es transformen en realitat.
And this is art. This is metaphor. But I think the point is that this is metaphor with teeth, and it's with those teeth that I want to propose today that we rethink a little bit about the role of contemporary math -- not just financial math, but math in general. That its transition from being something that we extract and derive from the world to something that actually starts to shape it -- the world around us and the world inside us. And it's specifically algorithms, which are basically the math that computers use to decide stuff. They acquire the sensibility of truth because they repeat over and over again, and they ossify and calcify, and they become real.
Estava pensant en això, ara fa dos anys en un vol transatlàntic perquè resulta que seia al costat d'un físic hongarès de la meva edat i estàvem xerrant de com era la vida durant la guerra freda pels físics hongaresos I li vaig dir, "I doncs, què fèieu?"
And I was thinking about this, of all places, on a transatlantic flight a couple of years ago, because I happened to be seated next to a Hungarian physicist about my age and we were talking about what life was like during the Cold War for physicists in Hungary. And I said, "So what were you doing?"
i ell va dir, "Bé, bàsicament, trencàvem l'Stealth."
And he said, "Well we were mostly breaking stealth."
I jo, "És una bona tasca. Interessant. Com funciona?" Per entendre això, cal que entengueu una mica com funciona l'Stealth. Això és una simplificació, però bé, bàsicament, no pots simplement travessar una senyal de radar amb 156 tones de metall al cel. No pot simplement desaparèixer. Però si poguessis agafar aquesta gran cosa, i la poguessis transformar en un milió de petites coses -- com un esbart d'ocells -- aleshores el radar que està buscant hauria de ser capaç de veure tots els esbarts d'ocells del cel. I, si ets un radar, no és massa bona idea.
And I said, "That's a good job. That's interesting. How does that work?" And to understand that, you have to understand a little bit about how stealth works. And so -- this is an over-simplification -- but basically, it's not like you can just pass a radar signal right through 156 tons of steel in the sky. It's not just going to disappear. But if you can take this big, massive thing, and you could turn it into a million little things -- something like a flock of birds -- well then the radar that's looking for that has to be able to see every flock of birds in the sky. And if you're a radar, that's a really bad job.
I va dir: "Si". Va dir: "Això si ets un radar. De manera que no vam fer servir un radar; vam construir una caixa negra que buscava senyals elèctriques, comunicacions electròniques. I quan veiem un esbart d'ocells que tenia comunicacions electròniques, sabíem que segurament tenia a veure amb els americans."
And he said, "Yeah." He said, "But that's if you're a radar. So we didn't use a radar; we built a black box that was looking for electrical signals, electronic communication. And whenever we saw a flock of birds that had electronic communication, we thought, 'Probably has something to do with the Americans.'"
I jo vaig dir, "Sí, molt bé. Per tant, acabes de destrossar 60 anys de recerca aeronàutica i després, què? Què vas fer quan vas créixer? I va dir, "Serveis financers" I jo vaig dir, "Oh." Perquè feia poc havia sortit a les notícies. I li vaig dir: "Com funciona?" I ell em va dir, "Bé, ara mateix hi ha uns 2000 físics a Wall Street, i jo sóc un d'ells." I jo vaig dir: "I com és la caixa negra a Wall Street?"
And I said, "Yeah. That's good. So you've effectively negated 60 years of aeronautic research. What's your act two? What do you do when you grow up?" And he said, "Well, financial services." And I said, "Oh." Because those had been in the news lately. And I said, "How does that work?" And he said, "Well there's 2,000 physicists on Wall Street now, and I'm one of them." And I said, "What's the black box for Wall Street?"
I va dir: "És graciós que em preguntis això, perquè, en realitat, es parla de comerç de caixa negra, o també comerç algo, comerç algorítmic." I el comerç algorítmic ha evolucionat, en part, perquè els comerciants institucionals tenen els mateixos problemes que les Forces Armades dels Estats Units tenien, és a dir, que estan movent aquestes accions -- ja sigui "Proctor & Gamble" o "Accenture", és igual -- estan movent un milió d'accions d'alguna cosa a través dels mercats. I si les mouen totes de cop, passa com al pòquer: apostar-ho tot d'entrada. És com ensenyar les teves cartes. Per tant necessiten una manera -- és a dir, algoritmes -- per dividir aquell gran moviment en un milió de petites transaccions. I la màgia, o l'horror, de la qüestió es que es poden fer servir les mateixes matemàtiques tant per trencar la gran cosa en un milió de petites coses com per buscar un milió de petites coses i recomposar una gran cosa i, d'aquesta manera, saber que està passant realment als mercats.
And he said, "It's funny you ask that, because it's actually called black box trading. And it's also sometimes called algo trading, algorithmic trading." And algorithmic trading evolved in part because institutional traders have the same problems that the United States Air Force had, which is that they're moving these positions -- whether it's Proctor & Gamble or Accenture, whatever -- they're moving a million shares of something through the market. And if they do that all at once, it's like playing poker and going all in right away. You just tip your hand. And so they have to find a way -- and they use algorithms to do this -- to break up that big thing into a million little transactions. And the magic and the horror of that is that the same math that you use to break up the big thing into a million little things can be used to find a million little things and sew them back together and figure out what's actually happening in the market.
Així que si us voleu fer una idea del que passa a la borsa ara mateix, podeu pensar en una colla d'algoritmes pensats bàsicament per amagar, i una altra colla d'algoritmes pensats per trobar i actuar. I tot això està molt bé i representa el 70 % de la borsa dels Estats Units el 70 % del sistema operatiu altrament dit la teva pensió, la teva hipoteca,
So if you need to have some image of what's happening in the stock market right now, what you can picture is a bunch of algorithms that are basically programmed to hide, and a bunch of algorithms that are programmed to go find them and act. And all of that's great, and it's fine. And that's 70 percent of the United States stock market, 70 percent of the operating system formerly known as your pension, your mortgage.
Què podria fallar? Podria passar tal com fa un any, quan el 9% del total del mercat simplement va desaparèixer durant 5 minuts, i n'hi van dir el flash crash de les 2:45. Tot de cop, el 9% simplement desapareix, i fins ara, ningú ha pogut aclarir que va passar, perquè ningú ho va manar, ningú ho va demanar. De fet, ningú tenia cap control sobre el que realment succeía. Tot el que tenien era, simplement una pantalla davant d'ells amb un munt de números, i un botó vermell que posava "Stop."
And what could go wrong? What could go wrong is that a year ago, nine percent of the entire market just disappears in five minutes, and they called it the Flash Crash of 2:45. All of a sudden, nine percent just goes away, and nobody to this day can even agree on what happened because nobody ordered it, nobody asked for it. Nobody had any control over what was actually happening. All they had was just a monitor in front of them that had the numbers on it and just a red button that said, "Stop."
El que està passant, es que escrivim coses, coses que ja no podem llegir Hem generat quelcom il·legible I ja no sabem realment què està passant en aquest món que hem construït. I això és el començament. Hi ha una companyia a Boston, Nanex, utilitza matemàtiques i màgia no sé exactament com, però agafen dades dels mercats i troben, de vegades, alguns d'aquests algoritmes. Quan els troben, els extreuen i els claven a la paret com papallones. I aleshores fan allò que tots hem fet quan ens hem trobat davant de grans quantitats de dades que no enteniem-- simplement, els hi posen un nom i una història. A aquest que van trobar, li van posar "Knife", el "Carnival". el "Boston Shuffler" "Twilight".
And that's the thing, is that we're writing things, we're writing these things that we can no longer read. And we've rendered something illegible, and we've lost the sense of what's actually happening in this world that we've made. And we're starting to make our way. There's a company in Boston called Nanex, and they use math and magic and I don't know what, and they reach into all the market data and they find, actually sometimes, some of these algorithms. And when they find them they pull them out and they pin them to the wall like butterflies. And they do what we've always done when confronted with huge amounts of data that we don't understand -- which is that they give them a name and a story. So this is one that they found, they called the Knife, the Carnival, the Boston Shuffler, Twilight.
El més graciós de tot és que, òbviament, aquests algoritmes no funcionen només als mercats. Aquest tipus de coses es poden trobar arreu, un cop se sap com buscar-les. Per exemple, aquí: aquest llibre sobre mosques que potser us heu estat mirant per Amazon. Potser el vàreu veure quan costava 1.700.000 dòlars No s'imprimeix -- encara... (Rialles) Si l'haguéssiu comprat a 1.700.000 hauria sigut una ganga, doncs al cap d'unes hores havia pujat fins als 23.600.000 dòlars. costs d'enviament apart. La pregunta és: Si ningú estava comprant ni venent res; què estava passant? Aquests tipus de comportament s'observen tant a Amazon com a Wall Street. I el que aquests comportaments són una conseqüència d'algoritmes en conflicte, algoritmes bloquejats en bucles, sense cap supervisió humana, sense la supervisió d'un adult que pugui dir: "Home, 1.700.000 dòlars ja és bastant."
And the gag is that, of course, these aren't just running through the market. You can find these kinds of things wherever you look, once you learn how to look for them. You can find it here: this book about flies that you may have been looking at on Amazon. You may have noticed it when its price started at 1.7 million dollars. It's out of print -- still ... (Laughter) If you had bought it at 1.7, it would have been a bargain. A few hours later, it had gone up to 23.6 million dollars, plus shipping and handling. And the question is: Nobody was buying or selling anything; what was happening? And you see this behavior on Amazon as surely as you see it on Wall Street. And when you see this kind of behavior, what you see is the evidence of algorithms in conflict, algorithms locked in loops with each other, without any human oversight, without any adult supervision to say, "Actually, 1.7 million is plenty."
(Rialles)
(Laughter)
Passa el mateix amb Netflix. Netflix ha fet servir diferents algoritmes al llarg dels anys: van començar amb Cinematch, i després n'han anat provant d'altres. com "Dinosaur Planet". "Gravity". i, actualment, "Pragmatic Chaos". El que aquest algoritme intenta, igual tots els altres, es el següent: Tracta d'entendre't a tu, el que hi ha a dins teu, per a ser capaç de recomanar-te quina pel·lícula t'agradaria veure -- i això és una tasca molt, molt difícil. Però que la tasca sigui difícil, i el fet que encara no sabem realment com resoldre-la no elimina les conseqüències d'usar "Pragmatic Chaos". Al cap i a la fi, aquest algoritme, i tots els altres que Netflix ha usat, determina el 60% de les pel·lícules que els clients acaben llogant. És a dir, que un tros de codi, que conté una sola idea sobre tu, és el responsable del 60% de les vendes.
And as with Amazon, so it is with Netflix. And so Netflix has gone through several different algorithms over the years. They started with Cinematch, and they've tried a bunch of others -- there's Dinosaur Planet; there's Gravity. They're using Pragmatic Chaos now. Pragmatic Chaos is, like all of Netflix algorithms, trying to do the same thing. It's trying to get a grasp on you, on the firmware inside the human skull, so that it can recommend what movie you might want to watch next -- which is a very, very difficult problem. But the difficulty of the problem and the fact that we don't really quite have it down, it doesn't take away from the effects Pragmatic Chaos has. Pragmatic Chaos, like all Netflix algorithms, determines, in the end, 60 percent of what movies end up being rented. So one piece of code with one idea about you is responsible for 60 percent of those movies.
Ara bé, què passaria si poguéssim puntuar aquestes pel·lícules abans de que existissin? Oi que seria molt pràctic? Bé, alguns científics del Regne Unit, experts en dades, estan a Hollywood, desenvolupant algoritmes per guions -- Hi ha una empresa anomenada Epagogix que, a partir del teu guió et poden dir, quantitativament, si serà una pel·lícula de 30 milions de dòlars o una de 200 milions. I això no és Google, no és informació, ni estadística financera: això és cultura. I el que heu vist avui aquí, o el que normalment no veieu és la física de la cultura. I si aquests algoritmes, igual que els de Wall Street, produissin un crash, com ho sabríem? quina pinta faria?
But what if you could rate those movies before they get made? Wouldn't that be handy? Well, a few data scientists from the U.K. are in Hollywood, and they have "story algorithms" -- a company called Epagogix. And you can run your script through there, and they can tell you, quantifiably, that that's a 30 million dollar movie or a 200 million dollar movie. And the thing is, is that this isn't Google. This isn't information. These aren't financial stats; this is culture. And what you see here, or what you don't really see normally, is that these are the physics of culture. And if these algorithms, like the algorithms on Wall Street, just crashed one day and went awry, how would we know? What would it look like?
També són a casa teva. Això són dos algoritmes competint pel teu saló Són dos robots netejadors diferents i tenen idees força diferents sobre que vol dir netejar Això es pot veure posant càmera lenta i engantxant-los unes llums. Són com una mena d'arquitectes secrets a la teva habitació. I la idea que la pròpia arquitectura està subjecte a l'optimització algorítmica no és gens descabellada. És ben real, i està passant al teu voltant.
And they're in your house. They're in your house. These are two algorithms competing for your living room. These are two different cleaning robots that have very different ideas about what clean means. And you can see it if you slow it down and attach lights to them, and they're sort of like secret architects in your bedroom. And the idea that architecture itself is somehow subject to algorithmic optimization is not far-fetched. It's super-real and it's happening around you.
Ho notareu, per exemple, si esteu en una d'aquestes caixes metàl·liques hermètiques un d'aquests ascensors d'última generació ascensor amb control de destinació, se'n diuen. Són aquests on has de prémer el botó del pis on vas abans d'entrar a l'ascensor. Utilitzen un algoritme d'empaquetament per grups. És a dir, que res de la bogeria de deixar que cadascú es fiqui a l'ascensor que vol: Tothom qui va al desè pis, ha d'agafar el segon ascensor, i tothom qui va al tercer pis ha d'agafar el cinquè ascensor. El problema, però, és que la gent s'espanta. S'esgarrifa. I no és difícil veure perquè passa això. Passa perquè a l'ascensor li falten certs elements bàsics, com els botons. (rialles) Coses que la gent fa servir. Tot el que té és l'indicador del pis i el boto vermell que diu: "STOP". I això és el que estem creant. Estem creant aquesta comunicació amb les màquines I fins on podem arribar, estirant aquest concepte? Doncs bastant, bastant lluny
You feel it most when you're in a sealed metal box, a new-style elevator; they're called destination-control elevators. These are the ones where you have to press what floor you're going to go to before you get in the elevator. And it uses what's called a bin-packing algorithm. So none of this mishegas of letting everybody go into whatever car they want. Everybody who wants to go to the 10th floor goes into car two, and everybody who wants to go to the third floor goes into car five. And the problem with that is that people freak out. People panic. And you see why. You see why. It's because the elevator is missing some important instrumentation, like the buttons. (Laughter) Like the things that people use. All it has is just the number that moves up or down and that red button that says, "Stop." And this is what we're designing for. We're designing for this machine dialect. And how far can you take that? How far can you take it? You can take it really, really far.
Deixeu-me que torni a Wall Street. Els algoritmes de Wall Street depenent, per sobre de tot, d'una característica: la velocitat. Operen en milisegons i microsegons Per que us en feu una idea, es triguen uns 500.000 microsegons a fer clic amb el ratolí. Però si ets un algoritme de Wall Street i vas 5 microsegons enrere estàs perdut. Per tant, si fossis un algoritme buscaries un arquitecte com el que jo vaig conèixer a Frankfurt que estava buidant un gratacels -- llençant tots els mobles i tot allò que les persones usen, i posant acer al terra per a poder posar-hi piles de servidors -- tot per tal que un algoritme pogués estar més a prop d'Internet.
So let me take it back to Wall Street. Because the algorithms of Wall Street are dependent on one quality above all else, which is speed. And they operate on milliseconds and microseconds. And just to give you a sense of what microseconds are, it takes you 500,000 microseconds just to click a mouse. But if you're a Wall Street algorithm and you're five microseconds behind, you're a loser. So if you were an algorithm, you'd look for an architect like the one that I met in Frankfurt who was hollowing out a skyscraper -- throwing out all the furniture, all the infrastructure for human use, and just running steel on the floors to get ready for the stacks of servers to go in -- all so an algorithm could get close to the Internet.
Potser penseu que Internet es un sistema distribuït. És distribuït, però distribuït des de certs llocs. A Nova York, per exemple, es distribueix des d'aquí: l'Hotel Carrier al carrer Hudson. Aquí és de on realment surten els cables cap a la ciutat I el que passa és que si estàs més lluny d'aquí, estàs uns microsegons enrere. I aquesta gent de Wall Street, Marco Polo i Cherokee Nation, van 8 microsegons endarrere respecte tota aquesta gent que van als edificis que es buiden al voltant de l'hotel Carrier. I això seguirà passant. Seguirem buidant edificis, ja que, centímetre a centímetre, lliura a lliura i dòlar a dòlar cap de nosaltres podria treure més profit d'aquest espai que el Boston Shuffler.
And you think of the Internet as this kind of distributed system. And of course, it is, but it's distributed from places. In New York, this is where it's distributed from: the Carrier Hotel located on Hudson Street. And this is really where the wires come right up into the city. And the reality is that the further away you are from that, you're a few microseconds behind every time. These guys down on Wall Street, Marco Polo and Cherokee Nation, they're eight microseconds behind all these guys going into the empty buildings being hollowed out up around the Carrier Hotel. And that's going to keep happening. We're going to keep hollowing them out, because you, inch for inch and pound for pound and dollar for dollar, none of you could squeeze revenue out of that space like the Boston Shuffler could.
Però si ens allunyem, si ens allunyem, veiem una rasa de 1300 quilòmetres de llargada entre Nova York i Chicago que ha estat construïda darrerament per una empresa anomenada Spread Networks. Això és un cable de fibra òptica estès entre aquestes dues ciutats. només per transmetre un senyal 37 vegades més ràpid que el clic d'un ratolí, només per aquests algoritmes: el Carnival i el Knife. I quan pensem això, que estem foradant els Estats Units amb dinamita i serres de roca per a que un algoritme pugui tancar un contracte tres microsegons més ràpid, tot en un marc de comunicacions que cap humà arribarà a conèixer, és una mena de destí manifest que sempre cercarà una frontera nova.
But if you zoom out, if you zoom out, you would see an 825-mile trench between New York City and Chicago that's been built over the last few years by a company called Spread Networks. This is a fiber optic cable that was laid between those two cities to just be able to traffic one signal 37 times faster than you can click a mouse -- just for these algorithms, just for the Carnival and the Knife. And when you think about this, that we're running through the United States with dynamite and rock saws so that an algorithm can close the deal three microseconds faster, all for a communications framework that no human will ever know, that's a kind of manifest destiny; and we'll always look for a new frontier.
Però encara tenim molta feina a fer. Tot això es teoria d'uns matemàtics del MIT. I, la veritat, és que no entenc gaire del que parlen. Es tracta de cons lluminosos y connexions quàntiques. i el cert és que no comprenc res d'això. Això sí, puc llegir aquest mapa. que diu que si intentem guanyar diners als mercats a on es troben els punts vermells, és a dir, a on es troba la gent, a on són les ciutats, haurem de posar els servidors a on estan els punts blaus per tal d'obtenir la màxima eficiència. I potser heu notat que la majoria de punts blaus es troben al mig de l'oceà. Així doncs, haurem de construir bombolles o plataformes. En realitat, anem a compartir l'aigua per tal d'extreure diners a l'aire perquè allí hi ha un futur brillant si som algoritmes.
Unfortunately, we have our work cut out for us. This is just theoretical. This is some mathematicians at MIT. And the truth is I don't really understand a lot of what they're talking about. It involves light cones and quantum entanglement, and I don't really understand any of that. But I can read this map, and what this map says is that, if you're trying to make money on the markets where the red dots are, that's where people are, where the cities are, you're going to have to put the servers where the blue dots are to do that most effectively. And the thing that you might have noticed about those blue dots is that a lot of them are in the middle of the ocean. So that's what we'll do: we'll build bubbles or something, or platforms. We'll actually part the water to pull money out of the air, because it's a bright future if you're an algorithm.
(Rialles)
(Laughter)
En realitat, els diners no són el que ens interessa més, sinó més aviat la motivació que porten els diners. El fet de transformar el mateix planeta amb aquesta mena d'eficiència algorítmica. Sota aquesta perspectiva tornem a veure les fotografies de Michael Najjar i ens adonem de que no són metafòriques, són profètiques. S'anticipen als efectes sísmics, terrestres de les matemàtiques que fem. I el paisatge sempre ha estat configurat per aquesta mena de col·laboració, estranya i difícil, entre la natura i el home. Però ara existeix aquesta tercera força coevolutiva: els algoritmes; el "Boston Shuffler", el "Carnival". I haurem de considerar-los com una part de la natura. I d'alguna, ho són.
And it's not the money that's so interesting actually. It's what the money motivates, that we're actually terraforming the Earth itself with this kind of algorithmic efficiency. And in that light, you go back and you look at Michael Najjar's photographs, and you realize that they're not metaphor, they're prophecy. They're prophecy for the kind of seismic, terrestrial effects of the math that we're making. And the landscape was always made by this sort of weird, uneasy collaboration between nature and man. But now there's this third co-evolutionary force: algorithms -- the Boston Shuffler, the Carnival. And we will have to understand those as nature, and in a way, they are.
Gràcies.
Thank you.
(Aplaudiment)
(Applause)